Re: backup online redo log file during a hot backup

On Jun 22, 11:04 am, Massa Batheli <mng..._at_gmail.com> wrote:
> On Jun 22, 1:50 pm, "fitzjarr..._at_cox.net" <fitzjarr..._at_cox.net> wrote:
>
>
>
>
>
> > On Jun 22, 12:00 pm, Massa Batheli <mng..._at_gmail.com> wrote:
>
> > > This might seem out of place to those who have religiously not been
> > > backing up online redo log files but unfortunately some of us have
> > > been doing that.
> > > Question out of ignorance
>
> > > If you do not back them up would they be recreated from the archived
> > > redo log files when the files are restored and the database started /
> > > recovered?...
>
> > > Thanks for your response as usual
>
> > > Massa
>
> > The active redo log is not, I repeat not, archived unless you
> > explicitly do so with 'alter system archive log current;' at which
> > time it's no longer active, therefore if you lose this you may have
> > lost vital information necessary for instance recovery; you WILL have
> > lost the ability of Oracle to function. Active redo logs cannot be
> > 'recreated' from any archived copy of the redo logs. You really need
> > to read this:
>
> >http://download-west.oracle.com/docs/cd/B19306_01/backup.102/b14191/r...
>
> > before you go any further.
>
> > David Fitzjarrell
>
> David
> Thanks for your very fast reply
>
> But that was not the point of the question.The point was do you have
> to make a copy of the online redo log files as well
> as the datafiles during a backup or the redo log files(not archived
> redo log files are left out) ?If they are omitted would the database
> come up OK
>
> Massa

Actually, you generally _don't_ want to backup the redo logs. If the database is up, you will have an inconsistent copy of a redo log, since Oracle is writing to one set of them. Hot backups are actually inconsistent copies of data files, which are made consistent during recovery by reading archived redo logs, which are consistent copies of the redo made when it is finished being used. But remember, you back up tablespaces, and then you can restore and recover them.

So think of your database moving right along, and you have to recover just one tablespace. The redo log that is current has some transactions in it for the other tablespaces. You restore the problem tablespaces files, then recover it by applying archived logs. If you had restored the redo logs, you would have overwritten your last current redo log that had the transactions for the other tablespaces.

Do an alter database backup controlfile to trace and look at what you get in your udump directory - that shows where redo logs come from.

Personally, I still make periodic cold backups, including all files and executables, which can be used for certain rapid recovery in disasters, and some odd other things. That's the only reason to backup redo logs, so that oracle can find something when it comes up restored on a new system. On an existing system, there are reuseable files already there. RMAN doesn't even let you backup redo logs, a duplicate database command will handle it for you.

If you can work through some basic manual standby databases, including bringing them up in read-only mode, that can also help you understand when redo files are needed.

jg

--
@home.com is bogus.
"CUSTOMER IMPACT: Arithmetic errors are very bad" - from old Oracle
bug.